On Tue, 16 Oct 2007 11:21:23 -0400 Pete Wyckoff <pw at osc.edu> wrote: > RDMA data packets are not padded up to 4 byte boundaries, unlike TCP. > Reasons for this can be found in doc/README.iser. > > Signed-off-by: Pete Wyckoff <pw at osc.edu> > --- > usr/iscsi/iscsid.c | 24 +++++++++++++++++------- > 1 files changed, 17 insertions(+), 7 deletions(-) > > diff --git a/usr/iscsi/iscsid.c b/usr/iscsi/iscsid.c > index 8b7b48e..6d5471e 100644 > --- a/usr/iscsi/iscsid.c > +++ b/usr/iscsi/iscsid.c > @@ -1157,7 +1157,9 @@ static int iscsi_target_cmd_queue(struct iscsi_task *task) > uint32_t len; > void *buf; > > - len = roundup(task->read_len, 4); > + len = task->read_len; > + if (!conn->tp->rdma) > + len = roundup(len, 4); I want to avoid doing something like this: if (conn->tp->rdma) do iser else do tcp Can you add 'pad_len' to iscsi_transport struct and do something like: len = roundup(task->read_len, conn->tp->pad_len); tcp sets it to PAD_WORD_LEN and iser sets it to 1. |